Data synthesis and statistical analysis

Using an established algorithm, we estimated and pooled individual patient data to obtain an overall estimate of survival and freedom from structural valve deterioration.¹⁴ Owing to the lack of Kaplan-Meier curves for stroke and atrial fibrillation, we combined and presented the study results as incidence rates per 100 patient years post-intervention for each outcome using Metaprop’s DerSimonian and Laird random effects model, with a binomial distribution to model within study variability or stabilise variances by applying Freeman-Tukey double arcsine transformation.¹⁵

We addressed statistical heterogeneity through consistency of point estimates and extent of overlap of confidence intervals. Heterogeneity was not assessed with I² statistics, as this is typically not useful in prognostic studies with a large sample size and resulting precise estimates.¹³ To identify potential sources of heterogeneity, we performed subgroup analyses for age, valve type, and risk of bias, specified a priori. We hypothesised higher adverse event rates in older patients, in studies that included only bioprosthetic valves versus studies that also included mechanical valves. We established age thresholds consistent with the requirements for Rapid Recommendations: study mean or median age of ≤65, 65 to <75, 75 to <85, and ≥85.⁶ For survival estimates, we used the log-rank test to compare survival across the different age groups. We defined a half weighted threshold age in which close to 50% of the total sample, within each age subgroup, is above and below this threshold. For these age thresholds, we obtained life expectancy estimates from the Social Security Administration of United States of America.¹¹ We compared our median survival estimates with life expectancy estimates of the US general population.

A two sided P value of 0.05 or less was considered statistically significant. Review Manager 5 and STATA^{16 17} provided software for statistical analyses, as well as forest plots and funnel plots.